Masked Face Recognition using Convolutional Neural Network with FaceNet
As an important and promising technology in recent years, face recognition is widely used for biometric identification and authorization. Under the influence of the COVID-19 epidemic, a number of countries require their residents to wear masks in public places, which greatly increases the probability of face recognition failure. In this study, a feasible method based on unified embedding for masked face recognition is proposed. The FaceNet model is used to extract facial features. the CASIA and Labeled Face in-the-Wild (LFW) datasets are used for model training. The extraction of target features by the model is ensured by data augmentation and manual addition of masks. The trained model can detect masks with an accuracy of over 97% and recognize faces in mask-wearing situations. The effectiveness of the model is further verified in a real-time face recognition application.